sarvamai 1.1.16-alpha.2 → 1.1.17

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (80) hide show
  1. package/dist/cjs/BaseClient.js +2 -2
  2. package/dist/cjs/api/resources/chat/client/Client.js +8 -0
  3. package/dist/cjs/api/resources/chat/client/requests/CreateChatCompletionV2Request.d.ts +4 -4
  4. package/dist/cjs/api/resources/decisions/client/Client.d.ts +12 -4
  5. package/dist/cjs/api/resources/decisions/client/Client.js +12 -4
  6. package/dist/cjs/api/resources/decisions/client/requests/SystemOneRequest.d.ts +1 -1
  7. package/dist/cjs/api/resources/docAi/client/Client.d.ts +6 -13
  8. package/dist/cjs/api/resources/docAi/client/Client.js +6 -22
  9. package/dist/cjs/api/resources/docAi/client/requests/DocAiExtractRequest.d.ts +3 -14
  10. package/dist/cjs/api/resources/docAi/client/requests/DocAiExtractRequest.js +0 -5
  11. package/dist/cjs/api/resources/docAi/client/requests/DocAiGetResultsRequest.d.ts +3 -2
  12. package/dist/cjs/api/resources/docAi/client/requests/DocAiParseRequest.d.ts +4 -15
  13. package/dist/cjs/api/resources/docAi/client/requests/DocAiParseRequest.js +1 -7
  14. package/dist/cjs/api/resources/docAi/types/DocAiGetResultsResponse.d.ts +1 -4
  15. package/dist/cjs/api/resources/docAi/types/Format.d.ts +6 -0
  16. package/dist/cjs/api/resources/docAi/types/Format.js +9 -0
  17. package/dist/cjs/api/resources/docAi/types/index.d.ts +1 -0
  18. package/dist/cjs/api/resources/docAi/types/index.js +1 -0
  19. package/dist/cjs/api/resources/responses/client/Client.js +4 -0
  20. package/dist/cjs/api/resources/responses/client/requests/ResponsesCreateRequest.d.ts +5 -1
  21. package/dist/cjs/api/resources/responses/client/requests/ResponsesCreateStreamRequest.d.ts +5 -1
  22. package/dist/cjs/api/types/Choice.d.ts +1 -1
  23. package/dist/cjs/api/types/CreateResponseV2Request.d.ts +5 -1
  24. package/dist/cjs/api/types/DocAiJobStatusResponse.d.ts +1 -1
  25. package/dist/cjs/api/types/DocAiParseResults.d.ts +6 -39
  26. package/dist/cjs/api/types/DocAiPresignUploadResponse.d.ts +1 -1
  27. package/dist/cjs/api/types/ModelObject.d.ts +10 -0
  28. package/dist/cjs/api/types/ModelObject.js +9 -0
  29. package/dist/cjs/api/types/ReasoningEffort.d.ts +2 -1
  30. package/dist/cjs/api/types/ReasoningEffort.js +2 -1
  31. package/dist/cjs/api/types/ResponseReasoning.d.ts +1 -1
  32. package/dist/cjs/api/types/ResponseReasoningEffort.d.ts +2 -1
  33. package/dist/cjs/api/types/ResponseReasoningEffort.js +2 -1
  34. package/dist/cjs/api/types/ResponsesModelIds.d.ts +7 -1
  35. package/dist/cjs/api/types/ResponsesModelIds.js +7 -1
  36. package/dist/cjs/api/types/V2ModelIds.d.ts +8 -2
  37. package/dist/cjs/api/types/V2ModelIds.js +8 -2
  38. package/dist/cjs/version.d.ts +1 -1
  39. package/dist/cjs/version.js +1 -1
  40. package/dist/esm/BaseClient.mjs +2 -2
  41. package/dist/esm/api/resources/chat/client/Client.mjs +8 -0
  42. package/dist/esm/api/resources/chat/client/requests/CreateChatCompletionV2Request.d.mts +4 -4
  43. package/dist/esm/api/resources/decisions/client/Client.d.mts +12 -4
  44. package/dist/esm/api/resources/decisions/client/Client.mjs +12 -4
  45. package/dist/esm/api/resources/decisions/client/requests/SystemOneRequest.d.mts +1 -1
  46. package/dist/esm/api/resources/docAi/client/Client.d.mts +6 -13
  47. package/dist/esm/api/resources/docAi/client/Client.mjs +6 -22
  48. package/dist/esm/api/resources/docAi/client/requests/DocAiExtractRequest.d.mts +3 -14
  49. package/dist/esm/api/resources/docAi/client/requests/DocAiExtractRequest.mjs +0 -5
  50. package/dist/esm/api/resources/docAi/client/requests/DocAiGetResultsRequest.d.mts +3 -2
  51. package/dist/esm/api/resources/docAi/client/requests/DocAiParseRequest.d.mts +4 -15
  52. package/dist/esm/api/resources/docAi/client/requests/DocAiParseRequest.mjs +1 -7
  53. package/dist/esm/api/resources/docAi/types/DocAiGetResultsResponse.d.mts +1 -4
  54. package/dist/esm/api/resources/docAi/types/Format.d.mts +6 -0
  55. package/dist/esm/api/resources/docAi/types/Format.mjs +6 -0
  56. package/dist/esm/api/resources/docAi/types/index.d.mts +1 -0
  57. package/dist/esm/api/resources/docAi/types/index.mjs +1 -0
  58. package/dist/esm/api/resources/responses/client/Client.mjs +4 -0
  59. package/dist/esm/api/resources/responses/client/requests/ResponsesCreateRequest.d.mts +5 -1
  60. package/dist/esm/api/resources/responses/client/requests/ResponsesCreateStreamRequest.d.mts +5 -1
  61. package/dist/esm/api/types/Choice.d.mts +1 -1
  62. package/dist/esm/api/types/CreateResponseV2Request.d.mts +5 -1
  63. package/dist/esm/api/types/DocAiJobStatusResponse.d.mts +1 -1
  64. package/dist/esm/api/types/DocAiParseResults.d.mts +6 -39
  65. package/dist/esm/api/types/DocAiPresignUploadResponse.d.mts +1 -1
  66. package/dist/esm/api/types/ModelObject.d.mts +10 -0
  67. package/dist/esm/api/types/ModelObject.mjs +8 -1
  68. package/dist/esm/api/types/ReasoningEffort.d.mts +2 -1
  69. package/dist/esm/api/types/ReasoningEffort.mjs +2 -1
  70. package/dist/esm/api/types/ResponseReasoning.d.mts +1 -1
  71. package/dist/esm/api/types/ResponseReasoningEffort.d.mts +2 -1
  72. package/dist/esm/api/types/ResponseReasoningEffort.mjs +2 -1
  73. package/dist/esm/api/types/ResponsesModelIds.d.mts +7 -1
  74. package/dist/esm/api/types/ResponsesModelIds.mjs +7 -1
  75. package/dist/esm/api/types/V2ModelIds.d.mts +8 -2
  76. package/dist/esm/api/types/V2ModelIds.mjs +8 -2
  77. package/dist/esm/version.d.mts +1 -1
  78. package/dist/esm/version.mjs +1 -1
  79. package/package.json +1 -1
  80. package/reference.md +19 -18
@@ -43,8 +43,8 @@ function normalizeClientOptions(options) {
43
43
  const headers = (0, headers_js_1.mergeHeaders)({
44
44
  "X-Fern-Language": "JavaScript",
45
45
  "X-Fern-SDK-Name": "sarvamai",
46
- "X-Fern-SDK-Version": "1.1.16-alpha.2",
47
- "User-Agent": "sarvamai/1.1.16-alpha.2",
46
+ "X-Fern-SDK-Version": "1.1.17",
47
+ "User-Agent": "sarvamai/1.1.17",
48
48
  "X-Fern-Runtime": core.RUNTIME.type,
49
49
  "X-Fern-Runtime-Version": core.RUNTIME.version,
50
50
  }, options === null || options === void 0 ? void 0 : options.headers);
@@ -341,8 +341,12 @@ class ChatClient {
341
341
  throw new SarvamAI.BadRequestError(body, rawResponse);
342
342
  case 401:
343
343
  throw new SarvamAI.UnauthorizedError(body, rawResponse);
344
+ case 402:
345
+ throw new SarvamAI.PaymentRequiredError(body, rawResponse);
344
346
  case 403:
345
347
  throw new SarvamAI.ForbiddenError(body, rawResponse);
348
+ case 404:
349
+ throw new SarvamAI.NotFoundError(body, rawResponse);
346
350
  case 413:
347
351
  throw new SarvamAI.ContentTooLargeError(body, rawResponse);
348
352
  case 422:
@@ -351,8 +355,12 @@ class ChatClient {
351
355
  throw new SarvamAI.TooManyRequestsError(body, rawResponse);
352
356
  case 500:
353
357
  throw new SarvamAI.InternalServerError(body, rawResponse);
358
+ case 502:
359
+ throw new SarvamAI.BadGatewayError(body, rawResponse);
354
360
  case 503:
355
361
  throw new SarvamAI.ServiceUnavailableError(body, rawResponse);
362
+ case 504:
363
+ throw new SarvamAI.GatewayTimeoutError(body, rawResponse);
356
364
  default:
357
365
  throw new errors.SarvamAIError({
358
366
  statusCode,
@@ -20,14 +20,14 @@ export interface CreateChatCompletionV2Request {
20
20
  temperature?: number;
21
21
  /** Nucleus sampling: greater than `0` and at most `1`. On `glm5.3`, `top_p: 0` is rejected. Prefer changing either `top_p` or `temperature`, not both. */
22
22
  top_p?: number;
23
- /** On reasoning models (`glm5.3`, `glm5.3-flash`, `gemma4`, `deepseekv4.1-flash-1017`, `deepseekv4-pro`, `kimi-k3.1-pro`): **`low`**, **`high`**, or **`max`**. Default when omitted is **`max`**, except **`high`** on `glm5.3-flash`, `deepseekv4.1-flash-1017`, and `deepseekv4-pro`, and no reasoning on `gemma4`. **`none`** turns thinking off on `deepseekv4.1-flash-1017` and `deepseekv4-pro`. `kimi-k3.1-pro` always thinks. Trace in `message.reasoning_content`; billed as completion tokens and counts toward `max_tokens`. On `gemma4`, reasoning is off unless this is set (`low`, `high`, or `max`). */
23
+ /** On reasoning models: **`low`**, **`high`**, or **`max`** on every reasoning model, and **`medium`** on `sarvam-105b`, `deepseekv4.1-flash-1017`, `deepseekv4.1-flash-1017-interactive`, `deepseekv4-pro`, `kimi-k3.1-pro`, and `qwen38-27b` (`kimi-k3.1-pro` runs `medium` as `high`). Default when omitted is **`max`** on `glm5.3` and `kimi-k3.1-pro`, **`high`** on `glm5.3-flash`, `deepseekv4.1-flash-1017`, `deepseekv4.1-flash-1017-interactive`, and `deepseekv4-pro`, and no reasoning on `gemma4`; `qwen38-27b` thinks by default. **`none`** turns thinking off on `deepseekv4.1-flash-1017`, `deepseekv4.1-flash-1017-interactive`, `deepseekv4-pro`, and `qwen38-27b`. `kimi-k3.1-pro` always thinks: `none`, `minimal`, and `xhigh` return `400`. Trace in `message.reasoning_content`; billed as completion tokens and counts toward `max_tokens`. On `gemma4`, reasoning is off unless this is set (`low`, `high`, or `max`). */
24
24
  reasoning_effort?: SarvamAI.ReasoningEffort;
25
25
  /** Maximum tokens to generate in the completion, including reasoning tokens on reasoning models. Default `2048`. Set explicitly when you need longer answers — a low limit can yield `finish_reason: "length"` with little or no visible `content` if reasoning consumes the budget. */
26
26
  max_tokens?: number;
27
27
  /** Optional. Default `false` — omit or `false` for one JSON response. If `true`, the response is streamed as server-sent events and ends with a `[DONE]` line. Use `stream_options.include_usage: true` to receive a final usage chunk (including `reasoning_tokens`). That chunk has an empty `choices` array — do not assume `choices[0]` is present on every event. */
28
28
  stream?: boolean;
29
29
  stop?: SarvamAI.StopConfiguration;
30
- /** Number of chat completion choices to generate per request. You are billed for tokens across all choices. Default `1`. On `glm5.3`, only `n: 1` is accepted; higher values return `400 invalid_request_error`. `n` > 1 needs `temperature` greater than `0` on every model; `temperature: 0` returns `400`. */
30
+ /** Number of chat completion choices. Only `1` is supported (the default); values above `1` return `400 invalid_request_error`. Send separate requests for more completions. */
31
31
  n?: number;
32
32
  /** Optional integer seed for OpenAI SDK compatibility. */
33
33
  seed?: number;
@@ -59,8 +59,8 @@ export interface CreateChatCompletionV2Request {
59
59
  parallel_tool_calls?: boolean;
60
60
  /** Only valid with `stream: true` (otherwise `400`). Must be a top-level field, not inside `extra_body`. Use `{"include_usage": true}` to receive token usage in the final SSE chunk. */
61
61
  stream_options?: SarvamAI.ChatStreamOptions;
62
- /** If `true`, return log probabilities for output tokens (use with `top_logprobs`). Supported on **`sarvam-105b`**, **`gemma4`**, **`glm5.3-flash`**, **`deepseekv4.1-flash-1017`**, **`deepseekv4-pro`**, and **`kimi-k3.1-pro`**. Not supported on **`glm5.3`** (`400` if enabled). */
62
+ /** If `true`, return log probabilities for output tokens (use with `top_logprobs`). Supported on **`sarvam-105b`**, **`gemma4`**, **`glm5.3-flash`**, **`deepseekv4.1-flash-1017`**, **`deepseekv4.1-flash-1017-interactive`**, **`deepseekv4-pro`**, and **`qwen38-27b`**. Not supported on **`glm5.3`** or **`kimi-k3.1-pro`** (`400` if enabled). */
63
63
  logprobs?: boolean;
64
- /** How many top token log probabilities to return at each position. Takes effect only when `logprobs` is `true`; without it, some models ignore the field and others return **`400`**. Same model support as `logprobs`: works on `sarvam-105b`, `gemma4`, `glm5.3-flash`, `deepseekv4.1-flash-1017`, `deepseekv4-pro`, and `kimi-k3.1-pro`; on `glm5.3`, omit entirely — do not send `0`, since many SDKs add it by default. */
64
+ /** How many top token log probabilities to return at each position. Takes effect only when `logprobs` is `true`; without it, some models ignore the field and others return **`400`**. Same model support as `logprobs`: works on `sarvam-105b`, `gemma4`, `glm5.3-flash`, `deepseekv4.1-flash-1017`, `deepseekv4.1-flash-1017-interactive`, `deepseekv4-pro`, and `qwen38-27b`; on `glm5.3` and `kimi-k3.1-pro`, omit entirely — do not send `0`, since many SDKs add it by default. */
65
65
  top_logprobs?: number;
66
66
  }
@@ -11,17 +11,25 @@ export declare class DecisionsClient {
11
11
  protected readonly _options: NormalizedClientOptionsWithAuth<DecisionsClient.Options>;
12
12
  constructor(options?: DecisionsClient.Options);
13
13
  /**
14
- * Answers a set of typed questions about one input and returns a calibrated probability for every allowed option. The model does not generate text, so there is nothing to parse. The order of `choice` options does not affect the answer.
14
+ * Answers a set of typed questions about one input and returns a calibrated probability for every allowed option. The model does not generate text, so there is nothing to parse.
15
+ *
16
+ * **Models:** `sarvam-prism` (serverless) and `clef` (Provisioned Throughput; also reads images).
15
17
  *
16
18
  * Requires `api-subscription-key`, or `Authorization: Bearer <key>`.
17
19
  *
18
20
  * **Question types:** `noul` (yes/no), `choice` (one of your options) and `score` (a level on your ordered scale).
19
21
  *
20
- * **Repeatability:** probabilities can vary slightly between identical requests; the top answer is stable.
22
+ * **Repeatability:** on `sarvam-prism`, probabilities can vary slightly between identical requests and the top answer is stable; the order of `choice` options does not affect the answer. On `clef`, identical requests return identical probabilities.
23
+ *
24
+ * **Images:** `clef` reads base64 images in `images`; image URLs return `400`. `sarvam-prism` refuses `images` (image URLs return `400`, inline images `422`).
25
+ *
26
+ * **Not supported:** streaming. `temperature` returns `422` on both models. `sarvam-prism` also refuses `max_tokens` and `stream` with `422`; `clef` accepts and ignores them. Unknown keys inside a question return `422`.
21
27
  *
22
- * **Not supported:** images and streaming. Generation fields such as `temperature`, `max_tokens`, `stream` and `images` are refused: image URLs return `400`, and the others return `422`. Unknown keys inside a question also return `422`.
28
+ * **Limits:**
29
+ * - `sarvam-prism`: 32,768 tokens per question (state + question + options), 2–255 options per `choice`, and 2–10 levels per `score`. If any one question breaks a limit, the whole request fails with `422`.
30
+ * - `clef`: 65,536 tokens per request (state, every question and its options, and images), 1–255 options per `choice`, and 1–10 levels per `score`. A longer request fails with `400`.
23
31
  *
24
- * **Limits:** 32,768 tokens per question (state + question + options), 2–255 options per `choice` and 2–10 levels per `score`. There is no cap on the number of questions. The state is not truncated, and if any one question breaks a limit the whole request fails with `422`.
32
+ * There is no cap on the number of questions, and the state is never truncated.
25
33
  *
26
34
  * **Timeout:** each request must finish within 60 seconds. A longer request fails with `503 model_overloaded`; split very large batches across several requests before retrying.
27
35
  *
@@ -56,17 +56,25 @@ class DecisionsClient {
56
56
  this._options = (0, BaseClient_js_1.normalizeClientOptionsWithAuth)(options);
57
57
  }
58
58
  /**
59
- * Answers a set of typed questions about one input and returns a calibrated probability for every allowed option. The model does not generate text, so there is nothing to parse. The order of `choice` options does not affect the answer.
59
+ * Answers a set of typed questions about one input and returns a calibrated probability for every allowed option. The model does not generate text, so there is nothing to parse.
60
+ *
61
+ * **Models:** `sarvam-prism` (serverless) and `clef` (Provisioned Throughput; also reads images).
60
62
  *
61
63
  * Requires `api-subscription-key`, or `Authorization: Bearer <key>`.
62
64
  *
63
65
  * **Question types:** `noul` (yes/no), `choice` (one of your options) and `score` (a level on your ordered scale).
64
66
  *
65
- * **Repeatability:** probabilities can vary slightly between identical requests; the top answer is stable.
67
+ * **Repeatability:** on `sarvam-prism`, probabilities can vary slightly between identical requests and the top answer is stable; the order of `choice` options does not affect the answer. On `clef`, identical requests return identical probabilities.
68
+ *
69
+ * **Images:** `clef` reads base64 images in `images`; image URLs return `400`. `sarvam-prism` refuses `images` (image URLs return `400`, inline images `422`).
70
+ *
71
+ * **Not supported:** streaming. `temperature` returns `422` on both models. `sarvam-prism` also refuses `max_tokens` and `stream` with `422`; `clef` accepts and ignores them. Unknown keys inside a question return `422`.
66
72
  *
67
- * **Not supported:** images and streaming. Generation fields such as `temperature`, `max_tokens`, `stream` and `images` are refused: image URLs return `400`, and the others return `422`. Unknown keys inside a question also return `422`.
73
+ * **Limits:**
74
+ * - `sarvam-prism`: 32,768 tokens per question (state + question + options), 2–255 options per `choice`, and 2–10 levels per `score`. If any one question breaks a limit, the whole request fails with `422`.
75
+ * - `clef`: 65,536 tokens per request (state, every question and its options, and images), 1–255 options per `choice`, and 1–10 levels per `score`. A longer request fails with `400`.
68
76
  *
69
- * **Limits:** 32,768 tokens per question (state + question + options), 2–255 options per `choice` and 2–10 levels per `score`. There is no cap on the number of questions. The state is not truncated, and if any one question breaks a limit the whole request fails with `422`.
77
+ * There is no cap on the number of questions, and the state is never truncated.
70
78
  *
71
79
  * **Timeout:** each request must finish within 60 seconds. A longer request fails with `503 model_overloaded`; split very large batches across several requests before retrying.
72
80
  *
@@ -20,7 +20,7 @@ import type * as SarvamAI from "../../../../index.js";
20
20
  * }
21
21
  */
22
22
  export interface SystemOneRequest {
23
- /** The decision model to use. Currently `sarvam-prism` (Sarvam Prism). */
23
+ /** The decision model to use. `sarvam-prism` (Sarvam Prism, serverless) or `clef` (Clef, on Provisioned Throughput; also accepts base64 `images`, up to 65,536 tokens per request). */
24
24
  model: string;
25
25
  /** The input to decide on: text, or an object or array that is rendered as compact JSON. Required; an empty string is allowed. It is never truncated: if any one question (state + question + options) exceeds 32,768 tokens the request fails with `422`. */
26
26
  state: SystemOneRequest.State;
@@ -16,7 +16,7 @@ export declare class DocAiClient {
16
16
  * **Flow:**
17
17
  * 1. Call this endpoint with the MIME type you will upload (e.g. `application/pdf`).
18
18
  * 2. Upload the file using the returned `method`, `url` and `headers` before `expires_at`.
19
- * 3. Pass the returned `upload_id` through the `upload_ids` field on a later parse, extract, classify, or split job request.
19
+ * 3. Pass the returned `upload_id` through the `upload_ids` field on a later parse or extract job request.
20
20
  *
21
21
  * When `content_type` is set, the upload's `Content-Type` header must match it.
22
22
  *
@@ -55,9 +55,7 @@ export declare class DocAiClient {
55
55
  *
56
56
  * @example
57
57
  * import { createReadStream } from "fs";
58
- * await client.docAi.parse({
59
- * language: "en-IN"
60
- * })
58
+ * await client.docAi.parse({})
61
59
  */
62
60
  parse(request: SarvamAI.DocAiParseRequest, requestOptions?: DocAiClient.RequestOptions): core.HttpResponsePromise<SarvamAI.DocAiStartJobResponse>;
63
61
  private __parse;
@@ -86,19 +84,16 @@ export declare class DocAiClient {
86
84
  * import { createReadStream } from "fs";
87
85
  * await client.docAi.extract({
88
86
  * schema: "{\"type\":\"object\",\"properties\":{\"policy_number\":{\"type\":\"string\",\"description\":\"Insurance policy number\"},\"insured_name\":{\"type\":\"string\",\"description\":\"Name of the insured person\"},\"sum_insured\":{\"type\":\"number\",\"description\":\"Total sum insured, in INR\"}}}",
89
- * output_format: "json",
90
- * mode: "basic"
87
+ * output_format: "json"
91
88
  * })
92
89
  */
93
90
  extract(request: SarvamAI.DocAiExtractRequest, requestOptions?: DocAiClient.RequestOptions): core.HttpResponsePromise<SarvamAI.DocAiStartJobResponse>;
94
91
  private __extract;
95
92
  /**
96
- * Creates a Classify job. Classify labels a document against the categories you define, or suggests a document type when you supply none.
93
+ * Creates a Classify job. Classify labels a document, or each page, against the categories you define, or suggests a document type when you supply none.
97
94
  *
98
95
  * **Input:** send `file` (PDF, PNG, JPEG or TIFF; repeat the field for up to 10 documents) or `upload_ids`, plus a JSON `request` with the options below. You can also send a JSON body with `input` set to `{"document_ref": "..."}` to classify a child document returned by Split.
99
96
  *
100
- * **Availability:** options are enabled per deployment. If your deployment does not support an option, the request fails with `400 INVALID_REQUEST` (for example `additional properties 'scope' not allowed`). Call `GET /doc-ai/v1/capabilities` to see what is enabled, and send only those options.
101
- *
102
97
  * **Processing:** the job runs asynchronously. Poll `GET /doc-ai/v1/job/{job_id}/status` until it is `completed`, then fetch `GET /doc-ai/v1/job/{job_id}/results`. Add `?include=details` to the results for inspected pages, provenance and evidence.
103
98
  *
104
99
  * **Review:** `review_required` tells you whether a person should check the answer. `confidence` can be `null`; it is a token-based score, not a measured probability.
@@ -127,9 +122,7 @@ export declare class DocAiClient {
127
122
  *
128
123
  * **Input:** send `file` (PDF, PNG, JPEG or TIFF; repeat the field for up to 10 documents) or `upload_ids`, plus a JSON `request`.
129
124
  *
130
- * **Availability:** options are enabled per deployment. If your deployment does not support an option, the request fails with `400 INVALID_REQUEST` (for example `additional properties 'scope' not allowed`). Call `GET /doc-ai/v1/capabilities` to see what is enabled, and send only those options.
131
- *
132
- * **Processing:** the job runs asynchronously. Poll `GET /doc-ai/v1/job/{job_id}/status` until it is `completed`, then fetch `GET /doc-ai/v1/job/{job_id}/results`. `completed` means processing finished, not that every segment passed review: check `review_required` on each segment.
125
+ * **Processing:** the job runs asynchronously. Poll `GET /doc-ai/v1/job/{job_id}/status` until it is `completed`, then fetch `GET /doc-ai/v1/job/{job_id}/results`. `completed` means processing finished, not that every segment passed review: check each segment's `review.required`.
133
126
  *
134
127
  * @param {SarvamAI.DocAiSplitRequest} request
135
128
  * @param {DocAiClient.RequestOptions} requestOptions - Request-specific configuration.
@@ -183,7 +176,7 @@ export declare class DocAiClient {
183
176
  * })
184
177
  *
185
178
  * @example
186
- * await client.docAi.getResults("01a1b292-8d3d-710e-974c-fe45aa5bc7ce", {
179
+ * await client.docAi.getResults("20260707_9f1c2b3a-4d5e-6f70-8a9b-c0d1e2f3a4b5", {
187
180
  * format: "json"
188
181
  * })
189
182
  *
@@ -62,7 +62,7 @@ class DocAiClient {
62
62
  * **Flow:**
63
63
  * 1. Call this endpoint with the MIME type you will upload (e.g. `application/pdf`).
64
64
  * 2. Upload the file using the returned `method`, `url` and `headers` before `expires_at`.
65
- * 3. Pass the returned `upload_id` through the `upload_ids` field on a later parse, extract, classify, or split job request.
65
+ * 3. Pass the returned `upload_id` through the `upload_ids` field on a later parse or extract job request.
66
66
  *
67
67
  * When `content_type` is set, the upload's `Content-Type` header must match it.
68
68
  *
@@ -147,9 +147,7 @@ class DocAiClient {
147
147
  *
148
148
  * @example
149
149
  * import { createReadStream } from "fs";
150
- * await client.docAi.parse({
151
- * language: "en-IN"
152
- * })
150
+ * await client.docAi.parse({})
153
151
  */
154
152
  parse(request, requestOptions) {
155
153
  return core.HttpResponsePromise.fromPromise(this.__parse(request, requestOptions));
@@ -169,9 +167,6 @@ class DocAiClient {
169
167
  if (request.language != null) {
170
168
  _body.append("language", request.language);
171
169
  }
172
- if (request.output_format != null) {
173
- _body.append("output_format", request.output_format);
174
- }
175
170
  if (request.content_type != null) {
176
171
  _body.append("content_type", request.content_type);
177
172
  }
@@ -248,8 +243,7 @@ class DocAiClient {
248
243
  * import { createReadStream } from "fs";
249
244
  * await client.docAi.extract({
250
245
  * schema: "{\"type\":\"object\",\"properties\":{\"policy_number\":{\"type\":\"string\",\"description\":\"Insurance policy number\"},\"insured_name\":{\"type\":\"string\",\"description\":\"Name of the insured person\"},\"sum_insured\":{\"type\":\"number\",\"description\":\"Total sum insured, in INR\"}}}",
251
- * output_format: "json",
252
- * mode: "basic"
246
+ * output_format: "json"
253
247
  * })
254
248
  */
255
249
  extract(request, requestOptions) {
@@ -273,15 +267,9 @@ class DocAiClient {
273
267
  if (request.config_id != null) {
274
268
  _body.append("config_id", request.config_id);
275
269
  }
276
- if (request.language != null) {
277
- _body.append("language", request.language);
278
- }
279
270
  if (request.output_format != null) {
280
271
  _body.append("output_format", request.output_format);
281
272
  }
282
- if (request.mode != null) {
283
- _body.append("mode", request.mode);
284
- }
285
273
  if (request.classification != null) {
286
274
  _body.append("classification", request.classification);
287
275
  }
@@ -334,12 +322,10 @@ class DocAiClient {
334
322
  });
335
323
  }
336
324
  /**
337
- * Creates a Classify job. Classify labels a document against the categories you define, or suggests a document type when you supply none.
325
+ * Creates a Classify job. Classify labels a document, or each page, against the categories you define, or suggests a document type when you supply none.
338
326
  *
339
327
  * **Input:** send `file` (PDF, PNG, JPEG or TIFF; repeat the field for up to 10 documents) or `upload_ids`, plus a JSON `request` with the options below. You can also send a JSON body with `input` set to `{"document_ref": "..."}` to classify a child document returned by Split.
340
328
  *
341
- * **Availability:** options are enabled per deployment. If your deployment does not support an option, the request fails with `400 INVALID_REQUEST` (for example `additional properties 'scope' not allowed`). Call `GET /doc-ai/v1/capabilities` to see what is enabled, and send only those options.
342
- *
343
329
  * **Processing:** the job runs asynchronously. Poll `GET /doc-ai/v1/job/{job_id}/status` until it is `completed`, then fetch `GET /doc-ai/v1/job/{job_id}/results`. Add `?include=details` to the results for inspected pages, provenance and evidence.
344
330
  *
345
331
  * **Review:** `review_required` tells you whether a person should check the answer. `confidence` can be `null`; it is a token-based score, not a measured probability.
@@ -425,9 +411,7 @@ class DocAiClient {
425
411
  *
426
412
  * **Input:** send `file` (PDF, PNG, JPEG or TIFF; repeat the field for up to 10 documents) or `upload_ids`, plus a JSON `request`.
427
413
  *
428
- * **Availability:** options are enabled per deployment. If your deployment does not support an option, the request fails with `400 INVALID_REQUEST` (for example `additional properties 'scope' not allowed`). Call `GET /doc-ai/v1/capabilities` to see what is enabled, and send only those options.
429
- *
430
- * **Processing:** the job runs asynchronously. Poll `GET /doc-ai/v1/job/{job_id}/status` until it is `completed`, then fetch `GET /doc-ai/v1/job/{job_id}/results`. `completed` means processing finished, not that every segment passed review: check `review_required` on each segment.
414
+ * **Processing:** the job runs asynchronously. Poll `GET /doc-ai/v1/job/{job_id}/status` until it is `completed`, then fetch `GET /doc-ai/v1/job/{job_id}/results`. `completed` means processing finished, not that every segment passed review: check each segment's `review.required`.
431
415
  *
432
416
  * @param {SarvamAI.DocAiSplitRequest} request
433
417
  * @param {DocAiClient.RequestOptions} requestOptions - Request-specific configuration.
@@ -573,7 +557,7 @@ class DocAiClient {
573
557
  * })
574
558
  *
575
559
  * @example
576
- * await client.docAi.getResults("01a1b292-8d3d-710e-974c-fe45aa5bc7ce", {
560
+ * await client.docAi.getResults("20260707_9f1c2b3a-4d5e-6f70-8a9b-c0d1e2f3a4b5", {
577
561
  * format: "json"
578
562
  * })
579
563
  *
@@ -3,8 +3,7 @@ import type * as core from "../../../../../core/index.js";
3
3
  * @example
4
4
  * {
5
5
  * schema: "{\"type\":\"object\",\"properties\":{\"policy_number\":{\"type\":\"string\",\"description\":\"Insurance policy number\"},\"insured_name\":{\"type\":\"string\",\"description\":\"Name of the insured person\"},\"sum_insured\":{\"type\":\"number\",\"description\":\"Total sum insured, in INR\"}}}",
6
- * output_format: "json",
7
- * mode: "basic"
6
+ * output_format: "json"
8
7
  * }
9
8
  */
10
9
  export interface DocAiExtractRequest {
@@ -14,17 +13,13 @@ export interface DocAiExtractRequest {
14
13
  upload_ids?: string;
15
14
  /** Inline extraction schema as a JSON string. The root must be `type: "object"` with non-empty `properties`; every field needs a `type` and a non-empty `description`. Supported types: `string`, `number`, `integer`, `boolean`, `object`, `array` (objects need `properties`, arrays need `items`); optional `enum`; maximum nesting depth 4. Exactly one of `schema` and `config_id` is required. */
16
15
  schema?: string;
17
- /** Saved extraction configuration ID. Exactly one of `schema` and `config_id` is required. To get your config_id, see [Reuse a schema](/api/api-guides-tutorials/document-intelligence/extract#reuse-a-schema). */
16
+ /** Saved extraction configuration ID. Exactly one of `schema` and `config_id` is required. To get your config_id, see [Reuse a schema](/api/api-guides-tutorials/document-intelligence/how-to/extract#reuse-a-schema). */
18
17
  config_id?: string;
19
- /** Language code of the document (BCP-47), e.g. `en-IN`, `hi-IN`. Always pass the correct code for best accuracy. */
20
- language?: string;
21
18
  /** Output format of the extraction results. */
22
19
  output_format?: DocAiExtractRequest.OutputFormat;
23
- /** Extraction mode. `basic` (default) runs the standard extraction pipeline; `advanced` runs the enhanced pipeline for complex documents. */
24
- mode?: DocAiExtractRequest.Mode;
25
20
  /** Enable document classification. Boolean sent as text: `true` or `false`. */
26
21
  classification?: string;
27
- /** Model to use for extraction. Leave unset to use the default. */
22
+ /** Model to use for extraction. */
28
23
  model?: string;
29
24
  }
30
25
  export declare namespace DocAiExtractRequest {
@@ -35,10 +30,4 @@ export declare namespace DocAiExtractRequest {
35
30
  readonly Xlsx: "xlsx";
36
31
  };
37
32
  type OutputFormat = (typeof OutputFormat)[keyof typeof OutputFormat];
38
- /** Extraction mode. `basic` (default) runs the standard extraction pipeline; `advanced` runs the enhanced pipeline for complex documents. */
39
- const Mode: {
40
- readonly Basic: "basic";
41
- readonly Advanced: "advanced";
42
- };
43
- type Mode = (typeof Mode)[keyof typeof Mode];
44
33
  }
@@ -10,9 +10,4 @@ var DocAiExtractRequest;
10
10
  Csv: "csv",
11
11
  Xlsx: "xlsx",
12
12
  };
13
- /** Extraction mode. `basic` (default) runs the standard extraction pipeline; `advanced` runs the enhanced pipeline for complex documents. */
14
- DocAiExtractRequest.Mode = {
15
- Basic: "basic",
16
- Advanced: "advanced",
17
- };
18
13
  })(DocAiExtractRequest || (exports.DocAiExtractRequest = DocAiExtractRequest = {}));
@@ -1,3 +1,4 @@
1
+ import type * as SarvamAI from "../../../../index.js";
1
2
  /**
2
3
  * @example
3
4
  * {
@@ -16,8 +17,8 @@
16
17
  * {}
17
18
  */
18
19
  export interface DocAiGetResultsRequest {
19
- /** Output format of the results. `json` (the default) returns the JSON results described below. To get the HTML or Markdown output of a Parse job, use `GET /doc-ai/v1/job/{job_id}/download-url`. */
20
- format?: "json";
20
+ /** Output format of the results. Choose `md` for Markdown, `html` for HTML, or `json` for structured JSON. If you omit `format`, the results are returned as `json` by default. */
21
+ format?: SarvamAI.Format;
21
22
  /** Set to `details` for Classify and Split jobs to add inspected pages, provenance and boundary details. Supporting evidence also needs `output.include_evidence` on the request. */
22
23
  include?: "details";
23
24
  }
@@ -1,9 +1,7 @@
1
1
  import type * as core from "../../../../../core/index.js";
2
2
  /**
3
3
  * @example
4
- * {
5
- * language: "en-IN"
6
- * }
4
+ * {}
7
5
  */
8
6
  export interface DocAiParseRequest {
9
7
  /** Document file(s) to parse. Repeatable multipart part. Exactly one of `file` and `upload_ids` must be provided. */
@@ -12,22 +10,13 @@ export interface DocAiParseRequest {
12
10
  upload_ids?: string;
13
11
  /** Language of the document in BCP-47 format (e.g. `hi-IN`, `en-IN`). Defaults to Indic: when omitted, the document is processed as an Indic-language document. To process an English document, set `language` to `en-IN`. Sarvam Vision supports 22 Indian languages plus English; see the [Sarvam Vision model page](/api/api-guides-tutorials/models/sarvam-vision) for the full list. */
14
12
  language?: string;
15
- /** Output format of the parsed document: `html` (default), `md` (Markdown), or `json` (structured blocks). Fetch the output with the download-url endpoint. */
16
- output_format?: DocAiParseRequest.OutputFormat;
17
- /** Nature of the document content: `printed` (typed or printed text), `handwritten`, or `mixed`. */
13
+ /** Nature of the document content. */
18
14
  content_type?: DocAiParseRequest.ContentType;
19
- /** Model to use for parsing. Leave unset to use the default. */
15
+ /** Model to use for digitisation. */
20
16
  model?: string;
21
17
  }
22
18
  export declare namespace DocAiParseRequest {
23
- /** Output format of the parsed document: `html` (default), `md` (Markdown), or `json` (structured blocks). Fetch the output with the download-url endpoint. */
24
- const OutputFormat: {
25
- readonly Html: "html";
26
- readonly Md: "md";
27
- readonly Json: "json";
28
- };
29
- type OutputFormat = (typeof OutputFormat)[keyof typeof OutputFormat];
30
- /** Nature of the document content: `printed` (typed or printed text), `handwritten`, or `mixed`. */
19
+ /** Nature of the document content. */
31
20
  const ContentType: {
32
21
  readonly Printed: "printed";
33
22
  readonly Handwritten: "handwritten";
@@ -4,13 +4,7 @@ Object.defineProperty(exports, "__esModule", { value: true });
4
4
  exports.DocAiParseRequest = void 0;
5
5
  var DocAiParseRequest;
6
6
  (function (DocAiParseRequest) {
7
- /** Output format of the parsed document: `html` (default), `md` (Markdown), or `json` (structured blocks). Fetch the output with the download-url endpoint. */
8
- DocAiParseRequest.OutputFormat = {
9
- Html: "html",
10
- Md: "md",
11
- Json: "json",
12
- };
13
- /** Nature of the document content: `printed` (typed or printed text), `handwritten`, or `mixed`. */
7
+ /** Nature of the document content. */
14
8
  DocAiParseRequest.ContentType = {
15
9
  Printed: "printed",
16
10
  Handwritten: "handwritten",
@@ -1,5 +1,5 @@
1
1
  import type * as SarvamAI from "../../../index.js";
2
- export type DocAiGetResultsResponse = SarvamAI.DocAiGetResultsResponse.Extract | SarvamAI.DocAiGetResultsResponse.Parse | SarvamAI.DocAiGetResultsResponse.Digitise | SarvamAI.DocAiGetResultsResponse.Classify | SarvamAI.DocAiGetResultsResponse.Split;
2
+ export type DocAiGetResultsResponse = SarvamAI.DocAiGetResultsResponse.Extract | SarvamAI.DocAiGetResultsResponse.Parse | SarvamAI.DocAiGetResultsResponse.Classify | SarvamAI.DocAiGetResultsResponse.Split;
3
3
  export declare namespace DocAiGetResultsResponse {
4
4
  interface Extract extends SarvamAI.DocAiExtractResults {
5
5
  type: "extract";
@@ -7,9 +7,6 @@ export declare namespace DocAiGetResultsResponse {
7
7
  interface Parse extends SarvamAI.DocAiParseResults {
8
8
  type: "parse";
9
9
  }
10
- interface Digitise extends SarvamAI.DocAiParseResults {
11
- type: "digitise";
12
- }
13
10
  interface Classify extends SarvamAI.DocAiClassifyResults {
14
11
  type: "classify";
15
12
  }
@@ -0,0 +1,6 @@
1
+ export declare const Format: {
2
+ readonly Md: "md";
3
+ readonly Html: "html";
4
+ readonly Json: "json";
5
+ };
6
+ export type Format = (typeof Format)[keyof typeof Format];
@@ -0,0 +1,9 @@
1
+ "use strict";
2
+ // This file was auto-generated by Fern from our API Definition.
3
+ Object.defineProperty(exports, "__esModule", { value: true });
4
+ exports.Format = void 0;
5
+ exports.Format = {
6
+ Md: "md",
7
+ Html: "html",
8
+ Json: "json",
9
+ };
@@ -1 +1,2 @@
1
1
  export * from "./DocAiGetResultsResponse.js";
2
+ export * from "./Format.js";
@@ -15,3 +15,4 @@ var __exportStar = (this && this.__exportStar) || function(m, exports) {
15
15
  };
16
16
  Object.defineProperty(exports, "__esModule", { value: true });
17
17
  __exportStar(require("./DocAiGetResultsResponse.js"), exports);
18
+ __exportStar(require("./Format.js"), exports);
@@ -175,6 +175,8 @@ class ResponsesClient {
175
175
  throw new SarvamAI.BadRequestError(body, rawResponse);
176
176
  case 401:
177
177
  throw new SarvamAI.UnauthorizedError(body, rawResponse);
178
+ case 402:
179
+ throw new SarvamAI.PaymentRequiredError(body, rawResponse);
178
180
  case 403:
179
181
  throw new SarvamAI.ForbiddenError(body, rawResponse);
180
182
  case 404:
@@ -221,6 +223,8 @@ class ResponsesClient {
221
223
  throw new SarvamAI.BadRequestError(_response.error.body, _response.rawResponse);
222
224
  case 401:
223
225
  throw new SarvamAI.UnauthorizedError(_response.error.body, _response.rawResponse);
226
+ case 402:
227
+ throw new SarvamAI.PaymentRequiredError(_response.error.body, _response.rawResponse);
224
228
  case 403:
225
229
  throw new SarvamAI.ForbiddenError(_response.error.body, _response.rawResponse);
226
230
  case 404:
@@ -19,7 +19,7 @@ export interface ResponsesCreateRequest {
19
19
  tools?: SarvamAI.ResponseFunctionTool[];
20
20
  /** Forwarded exactly as written: `none`, `auto`, `required`, `{ "type": "function", "name": "..." }`, or another object. */
21
21
  tool_choice?: ResponsesCreateRequest.ToolChoice;
22
- /** Optional reasoning config for `glm5.3`, `glm5.3-flash`, `gemma4`, `deepseekv4.1-flash-1017`, `deepseekv4-pro`, `kimi-k3.1-pro`. **`effort`:** on `glm5.3` and `glm5.3-flash`, `low` or `high` only (the literal string `max` returns `400`; omitting it gives `max` on `glm5.3` and `high` on `glm5.3-flash`); on `deepseekv4.1-flash-1017`, `deepseekv4-pro`, and `kimi-k3.1-pro`, `low`, `high`, or `max` all work directly. The trace is in the reasoning item's `content[0].text`; `summary` is filled only when `summary` is set. Optional **`summary`:** `auto` or `detailed`. */
22
+ /** Optional reasoning config for `glm5.3`, `glm5.3-flash`, `gemma4`, the DeepSeek models, `kimi-k3.1-pro`, and `qwen38-27b`. **`effort`:** on `glm5.3` and `glm5.3-flash`, `low` or `high` only (the literal string `max` returns `400`; omitting it gives `max` on `glm5.3` and `high` on `glm5.3-flash`); on `deepseekv4.1-flash-1017`, `deepseekv4.1-flash-1017-interactive`, and `deepseekv4-pro`, `low`, `high`, `max`, or `none` (thinking off); on `kimi-k3.1-pro`, `low`, `medium`, `high`, or `max` (thinking is always on); on `qwen38-27b`, `none` turns thinking off. The trace is in the reasoning item's `content[0].text`; `summary` is filled only when `summary` is set. Optional **`summary`:** `auto` or `detailed`. */
23
23
  reasoning?: SarvamAI.ResponseReasoning;
24
24
  /** Merged into the outbound body. Must not overwrite any modelled field (`400`). `max_tool_calls` and `prompt` are `400` here and in `extra_body` — not supported by any serving deployment. */
25
25
  extra_body?: Record<string, unknown>;
@@ -48,6 +48,10 @@ export interface ResponsesCreateRequest {
48
48
  background?: boolean;
49
49
  /** `400 invalid_request_error` — not supported by any serving deployment, top-level or via `extra_body`. */
50
50
  max_tool_calls?: number;
51
+ /** How many of the most likely tokens to return at each output position, with their log probabilities. Send it with `include: ["message.output_text.logprobs"]`; without `include`, `output[].content[].logprobs` comes back empty. `0` returns no logprobs. Supported on `sarvam-105b`, `glm5.3-flash`, `deepseekv4.1-flash-1017`, `deepseekv4.1-flash-1017-interactive`, `deepseekv4-pro`, and `qwen38-27b`; `glm5.3`, `gemma4`, and `kimi-k3.1-pro` return `400`. */
52
+ top_logprobs?: number;
53
+ /** Extra data to add to the response. `message.output_text.logprobs` returns token log probabilities (use with `top_logprobs`); `reasoning.encrypted_content` is accepted. An unknown value returns `400`. */
54
+ include?: string[];
51
55
  /** Accepts any additional properties */
52
56
  [key: string]: any;
53
57
  }
@@ -19,7 +19,7 @@ export interface ResponsesCreateStreamRequest {
19
19
  tools?: SarvamAI.ResponseFunctionTool[];
20
20
  /** Forwarded exactly as written: `none`, `auto`, `required`, `{ "type": "function", "name": "..." }`, or another object. */
21
21
  tool_choice?: ResponsesCreateStreamRequest.ToolChoice;
22
- /** Optional reasoning config for `glm5.3`, `glm5.3-flash`, `gemma4`, `deepseekv4.1-flash-1017`, `deepseekv4-pro`, `kimi-k3.1-pro`. **`effort`:** on `glm5.3` and `glm5.3-flash`, `low` or `high` only (the literal string `max` returns `400`; omitting it gives `max` on `glm5.3` and `high` on `glm5.3-flash`); on `deepseekv4.1-flash-1017`, `deepseekv4-pro`, and `kimi-k3.1-pro`, `low`, `high`, or `max` all work directly. The trace is in the reasoning item's `content[0].text`; `summary` is filled only when `summary` is set. Optional **`summary`:** `auto` or `detailed`. */
22
+ /** Optional reasoning config for `glm5.3`, `glm5.3-flash`, `gemma4`, the DeepSeek models, `kimi-k3.1-pro`, and `qwen38-27b`. **`effort`:** on `glm5.3` and `glm5.3-flash`, `low` or `high` only (the literal string `max` returns `400`; omitting it gives `max` on `glm5.3` and `high` on `glm5.3-flash`); on `deepseekv4.1-flash-1017`, `deepseekv4.1-flash-1017-interactive`, and `deepseekv4-pro`, `low`, `high`, `max`, or `none` (thinking off); on `kimi-k3.1-pro`, `low`, `medium`, `high`, or `max` (thinking is always on); on `qwen38-27b`, `none` turns thinking off. The trace is in the reasoning item's `content[0].text`; `summary` is filled only when `summary` is set. Optional **`summary`:** `auto` or `detailed`. */
23
23
  reasoning?: SarvamAI.ResponseReasoning;
24
24
  /** Merged into the outbound body. Must not overwrite any modelled field (`400`). `max_tool_calls` and `prompt` are `400` here and in `extra_body` — not supported by any serving deployment. */
25
25
  extra_body?: Record<string, unknown>;
@@ -48,6 +48,10 @@ export interface ResponsesCreateStreamRequest {
48
48
  background?: boolean;
49
49
  /** `400 invalid_request_error` — not supported by any serving deployment, top-level or via `extra_body`. */
50
50
  max_tool_calls?: number;
51
+ /** How many of the most likely tokens to return at each output position, with their log probabilities. Send it with `include: ["message.output_text.logprobs"]`; without `include`, `output[].content[].logprobs` comes back empty. `0` returns no logprobs. Supported on `sarvam-105b`, `glm5.3-flash`, `deepseekv4.1-flash-1017`, `deepseekv4.1-flash-1017-interactive`, `deepseekv4-pro`, and `qwen38-27b`; `glm5.3`, `gemma4`, and `kimi-k3.1-pro` return `400`. */
52
+ top_logprobs?: number;
53
+ /** Extra data to add to the response. `message.output_text.logprobs` returns token log probabilities (use with `top_logprobs`); `reasoning.encrypted_content` is accepted. An unknown value returns `400`. */
54
+ include?: string[];
51
55
  /** Accepts any additional properties */
52
56
  [key: string]: any;
53
57
  }
@@ -9,7 +9,7 @@ export interface Choice {
9
9
  finish_reason: SarvamAI.FinishReason;
10
10
  /** The index of the choice in the list of choices. */
11
11
  index: number;
12
- /** Token log probabilities when supported (`sarvam-105b`, `gemma4`, `glm5.3-flash`, `deepseekv4.1-flash-1017`, `deepseekv4-pro`, `kimi-k3.1-pro`). Always `null` on `glm5.3`, which doesn't support this field. */
12
+ /** Token log probabilities when supported (`sarvam-105b`, `gemma4`, `glm5.3-flash`, `deepseekv4.1-flash-1017`, `deepseekv4.1-flash-1017-interactive`, `deepseekv4-pro`, `qwen38-27b`). Not supported on `glm5.3` or `kimi-k3.1-pro`, which return `400` when `logprobs` is set. */
13
13
  logprobs?: Record<string, unknown> | undefined;
14
14
  message: SarvamAI.ChatCompletionResponseMessage;
15
15
  }
@@ -14,7 +14,7 @@ export interface CreateResponseV2Request {
14
14
  tools?: SarvamAI.ResponseFunctionTool[] | undefined;
15
15
  /** Forwarded exactly as written: `none`, `auto`, `required`, `{ "type": "function", "name": "..." }`, or another object. */
16
16
  tool_choice?: CreateResponseV2Request.ToolChoice | undefined;
17
- /** Optional reasoning config for `glm5.3`, `glm5.3-flash`, `gemma4`, `deepseekv4.1-flash-1017`, `deepseekv4-pro`, `kimi-k3.1-pro`. **`effort`:** on `glm5.3` and `glm5.3-flash`, `low` or `high` only (the literal string `max` returns `400`; omitting it gives `max` on `glm5.3` and `high` on `glm5.3-flash`); on `deepseekv4.1-flash-1017`, `deepseekv4-pro`, and `kimi-k3.1-pro`, `low`, `high`, or `max` all work directly. The trace is in the reasoning item's `content[0].text`; `summary` is filled only when `summary` is set. Optional **`summary`:** `auto` or `detailed`. */
17
+ /** Optional reasoning config for `glm5.3`, `glm5.3-flash`, `gemma4`, the DeepSeek models, `kimi-k3.1-pro`, and `qwen38-27b`. **`effort`:** on `glm5.3` and `glm5.3-flash`, `low` or `high` only (the literal string `max` returns `400`; omitting it gives `max` on `glm5.3` and `high` on `glm5.3-flash`); on `deepseekv4.1-flash-1017`, `deepseekv4.1-flash-1017-interactive`, and `deepseekv4-pro`, `low`, `high`, `max`, or `none` (thinking off); on `kimi-k3.1-pro`, `low`, `medium`, `high`, or `max` (thinking is always on); on `qwen38-27b`, `none` turns thinking off. The trace is in the reasoning item's `content[0].text`; `summary` is filled only when `summary` is set. Optional **`summary`:** `auto` or `detailed`. */
18
18
  reasoning?: SarvamAI.ResponseReasoning | undefined;
19
19
  /** Merged into the outbound body. Must not overwrite any modelled field (`400`). `max_tool_calls` and `prompt` are `400` here and in `extra_body` — not supported by any serving deployment. */
20
20
  extra_body?: Record<string, unknown> | undefined;
@@ -43,6 +43,10 @@ export interface CreateResponseV2Request {
43
43
  background?: boolean | undefined;
44
44
  /** `400 invalid_request_error` — not supported by any serving deployment, top-level or via `extra_body`. */
45
45
  max_tool_calls?: number | undefined;
46
+ /** How many of the most likely tokens to return at each output position, with their log probabilities. Send it with `include: ["message.output_text.logprobs"]`; without `include`, `output[].content[].logprobs` comes back empty. `0` returns no logprobs. Supported on `sarvam-105b`, `glm5.3-flash`, `deepseekv4.1-flash-1017`, `deepseekv4.1-flash-1017-interactive`, `deepseekv4-pro`, and `qwen38-27b`; `glm5.3`, `gemma4`, and `kimi-k3.1-pro` return `400`. */
47
+ top_logprobs?: number | undefined;
48
+ /** Extra data to add to the response. `message.output_text.logprobs` returns token log probabilities (use with `top_logprobs`); `reasoning.encrypted_content` is accepted. An unknown value returns `400`. */
49
+ include?: string[] | undefined;
46
50
  /** Accepts any additional properties */
47
51
  [key: string]: any;
48
52
  }
@@ -4,7 +4,7 @@ export interface DocAiJobStatusResponse {
4
4
  job_id: string;
5
5
  /** Current status of the job. Terminal statuses: `completed`, `partially_completed`, `failed`, `rejected`. */
6
6
  status: string;
7
- /** Job pipeline: `digitise` for Parse jobs (the pipeline Parse runs on) or `extract`. Classify and Split jobs report `type` instead. */
7
+ /** Job pipeline: `parse` or `extract`. Classify and Split jobs report `type` instead. */
8
8
  pipeline?: string | undefined;
9
9
  usage: SarvamAI.DocAiUsage;
10
10
  /** Job creation timestamp. */